Pith. sign in

REVIEW 4 major objections 6 minor 28 references

Incremental Gain Computation and Regulation of Discrete-time Positive Lur\'e Systems using Linear Programming

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read For discrete-time positive systems in Lur'e feedback with elementwise-bounded nonlinearities, incremental l1 and l-infinity gains can be upper-bounded by solving linear programs, and the l-infinity gain can be regulated by a…

desk verdict Solid ℓ1 result; ℓ∞ and synthesis have an invalid substitution in the proof that needs fixing before the paper is reliable. read the letter →

arxiv 2505.24386 v1 pith:LTAZAL4N submitted 2025-05-30 math.OC cs.SYeess.SY

classification math.OCcs.SYeess.SY MSC 93C5593D2590C0593B52
keywords positivesystemsLur'eincrementalgainl1l-infinitylinearprogrammingstate-feedbackcontrolelementwiseLipschitzbound
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper aims to show that for discrete-time positive systems in Lur'e feedback with memoryless, elementwise-bounded nonlinearities, meaningful robustness certificates come cheaply: upper bounds on incremental $\ell_1$ and $\ell_\infty$ gains can be computed by solving linear programs, not nonlinear or semidefinite programs. Finite incremental gains quantify how far two trajectories can drift apart under different inputs, which matters for supply-chain bullwhip effects, biological networks, and any setting where worst-case deviations between trajectories are the object of interest. The paper further claims that the $\ell_\infty$ gain can be regulated by a state-feedback controller recovered directly from one linear program. If these claims are correct, gain analysis and controller synthesis for this class of nonlinear positive systems become tractable with standard LP solvers.

What carries the argument

The load-bearing device is the elementwise Lipschitz bound $\Delta$ in Assumption A2, which turns the memoryless nonlinearity into a linear majorant: $|z_1-z_2|\leq \Delta C_1|\delta x|+\Delta F_1|\delta w|$. Combined with nonnegativity of the plant matrices (A1), this lets every inequality in the proof go through with absolute values and nonnegative matrix products, so the nonlinear Lur'e system is bounded above by a positive linear system with matrices $A+B_1\Delta C_1$ and $B_2+B_1\Delta F_1$. The incremental storage $V(\delta x)=v^\top|\delta x|$ for $\ell_1$ analysis and the weighted sup-norm comparison for $\ell_\infty$ analysis reduce the gain bound to two linear constraints in the variables $(v,\beta,k,r)$. For controller design, the substitution $Y=K\,\mathrm{diag}(v)$ removes the bilinear term $Kv$ and makes the synthesis condition convex, with the controller recovered as $K=Y\,\mathrm{diag}(1/v)$.

What would settle it

Take any instance satisfying A1 and A2 where Program (11) is feasible with upper bound $\gamma$, simulate the closed-loop system from two initial states with a worst-case pair of input sequences, and compute the empirical incremental $\ell_1$ ratio $\sum_t\|y^1_t-y^2_t\|_1/\sum_t\|w^1_t-w^2_t\|_1$; a ratio strictly larger than $\gamma$ (after accounting for numerical precision) would disprove Theorem 3's upper-bound claim.

Watch

Extended reading notes

Core claim

The paper's central claim is that under two assumptions — all system matrices nonnegative (A1), and a known matrix $\Delta\geq 0$ with $|f(t,\zeta_1)-f(t,\zeta_2)|\leq \Delta|\zeta_1-\zeta_2|$ elementwise (A2) — the incremental $\ell_1$ gain from $w$ to $y$ is upper-bounded by $\gamma$ whenever the linear program (11) is feasible (Theorem 3), and the incremental $\ell_\infty$ gain is upper-bounded by $\eta$ whenever the linear program (12) is feasible (Theorem 4). Theorem 5 extends the $\ell_\infty$ result to synthesis: if the linear program (15) is feasible, the affine state-feedback law $u=Kx+g$ with $K=Y\,\mathrm{diag}(1/v)$ regulates the closed-loop incremental $\ell_\infty$ gain to at most $\eta$. The proofs use an incremental storage function $V(\delta x)=v^\top|\delta x|$ for the $\ell_1$ case and a weighted sup-norm argument for the $\ell_\infty$ case, with the nonlinearity replaced by its elementwise Lipschitz majorant $\Delta$.

Load-bearing premise

Everything depends on knowing a matrix $\Delta$ whose entries correctly bound every component of the difference $f(t,\zeta_1)-f(t,\zeta_2)$ for all $t$ and all pairs $\zeta_1,\zeta_2$; if $\Delta$ is unknown, underestimated, or the nonlinearity is not elementwise Lipschitz, the linear-programming certificates are invalid or vacuous.

Editorial extensions

If this is right

  • For any two input sequences, the sum over time of the output deviation in the 1-norm is at most $\gamma$ times the sum of the input deviations; the analogous statement holds with the $\ell_\infty$ norm and bound $\eta$.
  • With no external input, a finite incremental $\ell_1$ or $\ell_\infty$ gain forces every pair of trajectories to converge to the same fixed point, or to diverge together at the same rate.
  • The $\ell_\infty$ gain of the closed loop can be regulated below a prescribed $\eta$ by an affine state-feedback controller obtained from a single LP, so synthesis remains convex.
  • Because all certificates are LPs, the analysis scales to larger state dimensions than semidefinite-programming alternatives, and infeasibility of the LP at a given $\Delta$ marks where the elementwise Lipschitz certificate stops being valid.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • An implication the paper leaves implicit is that the method's conservatism is governed entirely by the tightness of $\Delta$; any prior sector or Lipschitz data on $f$ could be converted into a smaller $\Delta$, yielding sharper gain bounds without changing the LP structure.
  • The paper's remark that $\ell_1$ synthesis leads to a bilinear term suggests an asymmetry worth testing: $\ell_\infty$ regulation is LP-convex while $\ell_1$ regulation likely needs local or alternating methods, so a practical controller might trade off the two norms.
  • A testable extension is to replace the static matrix $\Delta$ with a time-varying or state-dependent bound inside the same LP framework; the proof only uses the elementwise comparison, so time-varying majorants would likely preserve the theorems.
  • The numerical examples suggest that the infeasibility threshold of the LP (for instance $\tau=0.125$ in the Leslie example) can be used as a robustness margin; one could compare that threshold with brute-force worst-case simulations to measure conservatism.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The manuscript proposes linear-programming conditions for upper-bounding incremental ℓ1 and ℓ∞ gains of discrete-time positive Lur'e systems, and an LP for state-feedback regulation of the ℓ∞ gain. Theorem 3 gives the ℓ1 bound, Theorem 4 the ℓ∞ bound, and Theorem 5 the synthesis result. The paper includes numerical examples on a Leslie population model and a randomly generated unstable positive system, with MATLAB code made publicly available.

Significance. If the theorems are correct, the paper offers tractable convex certificates for incremental performance of a useful class of nonlinear positive systems, going beyond the existing continuous-time ℓ1 treatment in [13]. The ℓ1 proof in Appendix A.2 is a clean dissipativity argument, and the numerical experiments are reproducible. However, the proof of Theorem 4 in Appendix A.3 contains a load-bearing algebraic error, so the ℓ∞ analysis and the regulation result depending on it are not established as written. The intended ℓ∞ result is plausible and likely repairable via a comparison-system argument, but that argument is not in the manuscript.

major comments (4)
  1. [Appendix A.3, Eqs. (35)–(37)] The step from (35) and (36) to (37) is invalid. Equation (35) contains (M|δx_t|)⊙v with M=A+B1∆C1, while equation (36), obtained by left-multiplying (12a) by |δx_t|, contains |δx_t|⊙(Mv). For non-diagonal nonnegative M these are not equal; e.g., M=[[0.5,0.4],[0.1,0.2]], v=(1,1), y=(1,100) gives My⊙v=(40.5,20.1) but y⊙Mv=(0.9,30). Thus inequality (37), and everything derived from it through (43), is not established, so Theorem 4 is unproved as written. Theorem 5, which applies Theorem 4 to the closed loop, inherits the gap. A comparison-system proof using ξ_{t+1}=Mξ_t+E|δw_t| with |δx_t|≤ξ_t may repair the argument, but it is not present. Additionally, Eqs. (42)–(43) multiply p-dimensional vectors such as η1_p, r, and C2v elementwise by the e-dimensional vector |δm_t|, which is only meaningful when p=e; and the bound F2|δw_t|≤(F2 1_e)⊙|δm_t| does not follow from F2≥0—the correct bound is F2|δm_t|.
  2. [Section 3.1, Assumption A2] The declared dimension of ∆ is inconsistent with its use. Since f(t,·):R^q→R^d, the inequality in (10) requires ∆∈R^{d×q}, not ∆∈R^{q×d}. With the stated dimension, ∆|ζ1−ζ2| has the wrong size and the products B1∆C1 appearing in Theorems 3–5 are undefined. This should be corrected in the assumption statement and in any accompanying text.
  3. [Section 3.2, Eq. (11b); Theorem 1, Eq. (7b)] The second constraint of the ℓ1 LP is dimensionally inconsistent. In (11b), γ1_p is p×1, while F2^T 1_e is undefined because F2^T∈R^{e×p}, and (B2+B1∆F1)^T v is e×1. The proof's equation (31) shows the intended constraint is γ1_e − F2^T 1_p − (B2+B1∆F1)^T v = r with r∈R^e. The same problem appears in the borrowed Theorem 1, Eq. (7b). As printed, the LP in Theorem 3 cannot be implemented, so the ℓ1 result needs this correction.
  4. [Appendix A.4, Eqs. (45d) and (46)] The proof of Theorem 5 contains local but consequential typos. Constraint (45d) is stated as A+B3K+B1∆(C1+D1K)∈R^{q×n}, but this matrix is n×n; the intended condition is (C1+D1K)∈R^{q×n}, which is what appears in the final program (15d). In the list of equivalences (46), the third identity repeats the C1 line and should be (C2+D2K)v=(C2 diag(v)+D2Y)1_n. These should be fixed because Theorem 5 is a central contribution.
minor comments (6)
  1. [Appendix A.3, after Eq. (35)] The proof says "By constraint (11a)" but the relevant constraint is (12a).
  2. [Appendix A.2, after Eq. (31)] The proof says "Substitution of relations (7a) and (7b)" but should refer to (11a) and (11b).
  3. [Appendix A.3, Eq. (39)] The displayed definition of e^+ appears to simplify to 1 whenever Σ_j E_ij>0; the intended scaling is unclear and should be clarified.
  4. [Section 5.1.1, text after Eq. (17)] The sentence refers to "the nonlinearity (16)" but the nonlinearity is defined in (17).
  5. [Figure 1 caption] The caption reads "`1 `1" and should presumably be "ℓ1 ℓ∞".
  6. [Section 2.2.2, Eqs. (4) and (6)] The definitions of incremental ℓ∞ and ℓ1 gain do not specify the initial-condition convention. Without a statement that the two trajectories start from the same state (or that the denominator includes an initial-state term), the ratios are undefined for identical input pairs with different initial states.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the LP certificates are derived from stated assumptions, not fitted to data or renamed inputs.

full rationale

The paper's claimed derivation chain is theorem-based rather than circular. Theorems 3 and 4 provide sufficient linear-programming conditions under Assumptions A1 and A2; the decision variables (v, beta, k, r, Y) are existential certificates satisfying explicit inequalities, not parameters fitted to observed data and then reported as predictions. The proofs in Appendix A derive the dissipativity inequalities directly from the dynamics and the elementwise Lipschitz bound in A2, using standard lemmas (triangle inequality and monotonicity of multiplication by nonnegative matrices). Assumption A2 is a modeling premise stated before the theorems, and the LP constraints are constructed from it; no theorem assumes the conclusion it is proving. Theorem 5 applies Theorem 4 to the closed-loop system after the standard change of variables Y = K diag(v), and the nonnegativity constraints are preserved by this transformation; this is a self-contained convexification argument, not a citation-loaded uniqueness claim. The citations to prior work [13], [10], and [28] supply external tools or comparison formulations, but the supporting proofs are included and do not depend on the target results. The numerical examples verify the computed bounds by simulation rather than using the simulation data to produce the gains. Even if the proof of Theorem 4 contains a substitution error (M|delta x| composed with v versus |delta x| composed with Mv), that is a correctness gap, not circularity: it does not make the claimed bound equivalent to its input by construction. Accordingly, the appropriate circularity finding is a clean non-finding with score 0.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

No new physical entities or fitted constants are introduced. The optimization variables v, beta, k, r, Y are existential certificates produced by the LPs, not data-fitted parameters. The only assumptions are structural positivity and Lipschitz bounds.

assumptions (5)
  • domain assumption A1: (A, B1, B2, C1, F1, F2) all have nonnegative entries
    Ensures the underlying linear system is positive; used in Lemma 2 and in all proofs to move absolute values through nonnegative matrices.
  • domain assumption A2: there exists a known Delta with |f(t,zeta1)-f(t,zeta2)| <= Delta|zeta1-zeta2| elementwise
    Load-bearing Lipschitz majorization that replaces the nonlinearity by a linear bound; all LPs in Sections 3 and 4 depend on Delta.
  • domain assumption A3: (B3, D1, D2) have nonnegative entries
    Ensures the control input preserves positivity of the closed-loop system; required for the synthesis result in Theorem 5.
  • standard math Standard positive-systems stability equivalences from [11], [16], [17]
    Used in Section 2.2.2 to characterize Schur stability of positive linear systems via positive vectors and diagonal Lyapunov functions.
  • standard math Lemma 1 (triangle inequality) and Lemma 2 (monotonicity of nonnegative matrix products)
    Proved in Appendix A.1 and used throughout the incremental gain proofs to bound absolute values of sums.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Incremental Gain Computation and Regulation of Discrete-time Positive Lur\'e Systems using Linear Programming." pith.science (2026). https://pith.science/paper/LTAZAL4N

@misc{pith2026250524386,
  author       = {Pith},
  title        = {Pith review of: Incremental Gain Computation and Regulation of Discrete-time Positive Lur\'e Systems using Linear Programming},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LTAZAL4N}},
  note         = {Machine review of arXiv:2505.24386}
}
abstract

This work approaches the problem of computing incremental $\ell_1$ and $\ell_\infty$ gains for discrete-time positive systems in \lure feedback with static memoryless nonlinearities, and regulating the $\ell_\infty$ gain through the design of a state-feedback controller. Finite incremental gains provide a quantitative measure of robustness for trajectories, and will ensure that all pairs of trajectories will converge to a fixed point or will diverge together in the absence of an applied input. Upper-bounds on these incremental gains can be computed through linear programming. Computation and regulation of the $\ell_1$ and $\ell_\infty$ incremental gains are verified by numerical examples.

Figures

Figures reproduced from arXiv: 2505.24386 by the authors.

Figure 1
Figure 1. reports upper bounds on the incremental gains ℓ1 and ℓ∞ between w → y for the Leslie system. All programs become infeasible at τ = 0.125, and the reported upper bound on the gain is thus ∞. 0 0.02 0.04 0.06 0.08 0.1 / 100 101 102 I n c r e m e n t al G ain `1 `1 [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Simulation of Leslie system starting at x0 = 15 5.2 System Control The second example involves a randomly generated positive linear system with n = 10 states, m = 4 controlled inputs, e = 2 external inputs, d = 2 Lur´e inputs, and p = 2 outputs. The uncontrolled system is unstable with a spectral radius of ρ(A) = 1.1888 [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. ℓ∞ bounds and observed values of the output for the Leslie system A.1 Technical Lemmas The following technical lemmas will be employed in proving Theorems 3, 4, and 5. The statements and proofs of these preliminary lemmas are enclosed for completeness. Lemma 1. Let c, z ∈ R n be vectors. Then |c ⊤z| ≤ |c| ⊤|z|. Proof. By the triangle inequality, |c ⊤z| ≤ Pn i=1|cizi|. The c terms can be pulled out by homogeneity, le… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Suboptimal regulation of the ℓ∞ norm The state transition has the relation of: v ⊤(x 1 t+1 − x 2 t+1) = (21) v ⊤A(x 1 t − x 2 t ) + v ⊤B1(z 1 t − z 2 t ) + v ⊤B2(w 1 t − w 2 t ) which consequently implies that |v ⊤(x 1 t+1 − x 2 t+1)| = (22) |v ⊤A(x 1 t − x 2 t ) + v ⊤…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

28 extracted references · 26 canonical work pages

  1. [13]

    A linear dissipativity approach to incremental input-to-state stability for a class of positive lur’e systems,

    V. Piengeon and C. Guiver, “A linear dissipativity approach to incremental input-to-state stability for a class of positive lur’e systems,”International Journal of Control, pp. 1–14, 2024

  2. [1]

    Compartmental system analysis: State of the art,

    R. F. Brown, “Compartmental system analysis: State of the art,”IEEE Trans. Biomed. Eng., no. 1, pp. 1–11, 2007

  3. [2]

    A positive systems model of TCP-like congestion control: asymptotic results,

    R. Shorten, F. Wirth, and D. Leith, “A positive systems model of TCP-like congestion control: asymptotic results,”IEEE/ACM transactions on networking, vol. 14, no. 3, pp. 616–629, 2006

  4. [3]

    Mitigating Transient Bullwhip Effects Under Imperfect Demand Forecasts

    S. H. Li and F. D¨ orfler, “Mitigating transient bullwhip effects under imperfect demand forecasts,” arXiv preprint arXiv:2404.01090, 2024

  5. [4]

    On contraction analysis for non-linear systems,

    W. Lohmiller and J.-J. E. Slotine, “On contraction analysis for non-linear systems,”Automatica, vol. 34, no. 6, pp. 683–696, 1998

  6. [5]

    C. A. Desoer and M. Vidyasagar,Feedback systems: input-output properties. SIAM, 2009

  7. [6]

    Polyhedral Estimation of L-1 and L-infinity Incremental Gains of Nonlinear Systems

    D. Kousoulidis and F. Forni, “Polyhedral estimation of l-1 and l-infinity incremental gains of non- linear systems,”arXiv preprint arXiv:2207.04334, 2022

  8. [7]

    Incremental positivity non-preservation by stability multipliers,

    V. Kulkarni and M. Safonov, “Incremental positivity non-preservation by stability multipliers,” in Proceedings of the 40th IEEE Conference on Decision and Control (Cat. No.01CH37228), vol. 1, pp. 33–38 vol.1, 2001

Show all 28 references
  1. [8]

    Data-driven representations of conical, convex, and affine behaviors,

    A. Padoan, F. D¨ orfler, and J. Lygeros, “Data-driven representations of conical, convex, and affine behaviors,” in2023 62nd IEEE Conference on Decision and Control (CDC), pp. 596–601, IEEE, 2023

  2. [9]

    L 1 gain analysis of linear positive systems and its appli- cation,

    Y. Ebihara, D. Peaucelle, and D. Arzelier, “L 1 gain analysis of linear positive systems and its appli- cation,” in2011 50th IEEE Conference on Decision and Control and European Control Conference, pp. 4029–4034, IEEE, 2011

  3. [10]

    Robust stability and stabilization of uncertain linear positive systems via integral linear constraints:l 1-gain andl ∞-gain characterization,

    C. Briat, “Robust stability and stabilization of uncertain linear positive systems via integral linear constraints:l 1-gain andl ∞-gain characterization,”International Journal of Robust and Nonlinear Control, vol. 23, no. 17, pp. 1932–1954, 2013

  4. [11]

    Farina and S

    L. Farina and S. Rinaldi,Positive Linear Systems: Theory and Applications, vol. 50. John Wiley & Sons, 2000

  5. [12]

    On the Kalman-Yakubovich-Popov Lemma for Positive Systems,

    A. Rantzer, “On the Kalman-Yakubovich-Popov Lemma for Positive Systems,”IEEE Transactions on Automatic Control, vol. 61, no. 5, pp. 1346–1349, 2016

  6. [14]

    The presence of a zero in an integer linear recurrent sequence is NP-hard to decide,

    V. D. Blondel and N. Portier, “The presence of a zero in an integer linear recurrent sequence is NP-hard to decide,”Linear algebra and its Applications, vol. 351, pp. 91–98, 2002

  7. [15]

    Identification of externally positive systems,

    C. Grussler, J. Umenberger, and I. R. Manchester, “Identification of externally positive systems,” in2017 IEEE 56th Annual Conference on Decision and Control (CDC), pp. 6549–6554, 2017

  8. [16]

    Kaczorek,Positive 1D and 2D systems

    T. Kaczorek,Positive 1D and 2D systems. Springer Science & Business Media, 2012. 12

  9. [17]

    A Tutorial on Positive Systems and Large Scale Control,

    A. Rantzer and M. E. Valcher, “A Tutorial on Positive Systems and Large Scale Control,” in2018 IEEE Conference on Decision and Control (CDC), pp. 3686–3697, IEEE, 2018

  10. [18]

    l 1 gain analysis of linear positive systems and its appli- cation,

    Y. Ebihara, D. Peaucelle, and D. Arzelier, “l 1 gain analysis of linear positive systems and its appli- cation,” in2011 50th IEEE Conference on Decision and Control and European Control Conference, pp. 4029–4034, 2011

  11. [19]

    Performance optimization over positivel ∞ cones,

    M. Naghnaeian and P. G. Voulgaris, “Performance optimization over positivel ∞ cones,” in2014 American Control Conference, pp. 5645–5650, 2014

  12. [20]

    ℓ1-induced norm and controller synthesis of positive systems,

    X. Chen, J. Lam, P. Li, and Z. Shu, “ℓ1-induced norm and controller synthesis of positive systems,” Automatica, vol. 49, no. 5, pp. 1377–1385, 2013

  13. [21]

    YALMIP : a toolbox for modeling and optimization in MATLAB,

    J. Lofberg, “YALMIP : a toolbox for modeling and optimization in MATLAB,” inICRA (IEEE Cat. No.04CH37508), pp. 284–289, 2004

  14. [22]

    ApS,The MOSEK optimization toolbox for MATLAB manual

    M. ApS,The MOSEK optimization toolbox for MATLAB manual. Version 9.2., 2020

  15. [23]

    On the use of matrices in certain population mathematics,

    P. H. Leslie, “On the use of matrices in certain population mathematics,”Biometrika, vol. 33, no. 3, pp. 183–212, 1945

  16. [24]

    L2+ Induced Norm Analysis of Continuous-Time LTI Systems Using Positive Filters and Copositive Programming,

    Y. Ebihara, H. Motooka, H. Waki, N. Sebe, V. Magron, D. Peaucelle, and S. Tarbouriech, “L2+ Induced Norm Analysis of Continuous-Time LTI Systems Using Positive Filters and Copositive Programming,” in2022 European Control Conference (ECC), pp. 1385–1390, 2022

  17. [25]

    Optimization algorithm synthesis based on integral quadratic constraints: A tutorial,

    C. W. Scherer, C. Ebenbauer, and T. Holicki, “Optimization algorithm synthesis based on integral quadratic constraints: A tutorial,” in2023 62nd IEEE Conference on Decision and Control (CDC), pp. 2995–3002, IEEE, 2023

  18. [26]

    Superstable control systems,

    B. Polyak, M. Sznaier, P. Shcherbakov, and M. Halpern, “Superstable control systems,”IF AC Proceedings Volumes, vol. 35, no. 1, pp. 199–204, 2002

  19. [27]

    S. Boyd, L. El Ghaoui, E. Feron, and V. Balakrishnan,Linear matrix inequalities in system and control theory. SIAM, 1994

  20. [28]

    Performance optimization over positivel∞ cones,

    M. Naghnaeian and P. G. Voulgaris, “Performance optimization over positivel∞ cones,”Automatica, vol. 80, pp. 177–188, 2017. 13

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.